ArchitectOS v79.0: An Offline-First, Geometry-Locked Computational Architecture and Local AI Orchestration Platform
Contemporary generative AI tools for architecture predominantly rely on cloud-hosted, prompt-driven diffusion models that lack deterministic geometric fidelity, local regulatory awareness, and data privacy. This paper presents ArchitectOS (v79.0), an offline-first computational architecture platform designed to orchestrate parametric multi-floor plan synthesis, 2D/3D CAD and BIM processing (DXF, DWG, IFC4, OBJ, GLB), and locally executed generative rendering. Unlike text-only image generators, ArchitectOS enforces a geometry-before-pixels paradigm: natural-language architectural briefs (with native Persian language support) and site surveys (Krocki) are first compiled into typed structural geometry (BuildingProjectResultV3), which subsequently drives depth- and line-art-conditioned local diffusion workflows via ComfyUI and local vision/language models via Ollama. We detail the system's dual-process Flask architecture, contract-based orchestration gates, fail-closed execution boundaries, and dual-source regulatory provenance framework. Based on a repository-grounded technical audit, we report verified release-acceptance results (77/77 passing architectural and isolation checks), scoped offline security test baselines, and an explicit account of current engineering limitations.
Authors
- Mahdi Ranjbar (ORCID: https://orcid.org/0000-0003-4223-4780)
Institutions
- Islamic Azad University Bandar Abbas (IR)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-20
- DOI
- https://doi.org/10.5281/zenodo.24407073
- Primary Topic
- Architecture and Computational Design
- Type
- article
- Field-Weighted Citation Impact
- 0.00